The rise of artificial intelligence and automation is fueling anxiety about the replacementof workers with robots and digital technologies. Relying upon a (country-sector-year) constructed measure of robotic capital (RK), we study the extent of complementarity/substitutabilitybetween robots and workers at different skill levels (i.e., high-, medium- and low-skilled workers). The analysis points to a higher elasticity of substitution (EoS) - i.e., higher substitutability - between RK and unskilled labor, compared to skilled labor. Furthermore, we find evidence of polarizing effects, according to which middle-skilled workers, typically employed in intermediate routine and/or codifiable tasks, are the most vulnerable to robotization. Results turn out to be robust to using different: i) definitions of EoS; ii) computations of RK; iii) samples of countries and industries (WIOD vs EU KLEMS data); iv) skill grouping.